diff --git a/.github/workflows/release.yaml b/.github/workflows/release.yaml
index bd1cff83..b2139b56 100644
--- a/.github/workflows/release.yaml
+++ b/.github/workflows/release.yaml
@@ -88,7 +88,7 @@ jobs:
           go install google.golang.org/protobuf/cmd/protoc-gen-go@v1.34.0
           export PATH=$PATH:$GOPATH/bin
           export PATH=/usr/local/cuda/bin:$PATH
-          GO_TAGS=p2p GOOS=linux GOARCH=arm64 CMAKE_ARGS="-DCMAKE_C_COMPILER=aarch64-linux-gnu-gcc -DCMAKE_CXX_COMPILER=aarch64-linux-gnu-g++" DIST_SKIP_HIPBLAS=true make dist
+          GO_TAGS=p2p GOOS=linux GOARCH=arm64 CMAKE_ARGS="-DCMAKE_C_COMPILER=aarch64-linux-gnu-gcc -DCMAKE_CXX_COMPILER=aarch64-linux-gnu-g++" make dist-aarch64
       - uses: actions/upload-artifact@v4
         with:
           name: LocalAI-linux-aarch64
diff --git a/Makefile b/Makefile
index d57c37c8..41bef790 100644
--- a/Makefile
+++ b/Makefile
@@ -324,13 +324,24 @@ build-api:
 dist:
 	STATIC=true $(MAKE) backend-assets/grpc/llama-cpp-avx2
 ifeq ($(OS),Darwin)
-	$(info ${GREEN}I Skip CUDA build on MacOS${RESET})
+	$(info ${GREEN}I Skip CUDA/hipblas build on MacOS${RESET})
 else
 	$(MAKE) backend-assets/grpc/llama-cpp-cuda
-ifneq ($(DIST_SKIP_HIPBLAS),true)
 	$(MAKE) backend-assets/grpc/llama-cpp-hipblas
 endif
+	$(MAKE) build
+	mkdir -p release
+# if BUILD_ID is empty, then we don't append it to the binary name
+ifeq ($(BUILD_ID),)
+	cp $(BINARY_NAME) release/$(BINARY_NAME)-$(OS)-$(ARCH)
+	shasum -a 256 release/$(BINARY_NAME)-$(OS)-$(ARCH) > release/$(BINARY_NAME)-$(OS)-$(ARCH).sha256
+else
+	cp $(BINARY_NAME) release/$(BINARY_NAME)-$(BUILD_ID)-$(OS)-$(ARCH)
+	shasum -a 256 release/$(BINARY_NAME)-$(BUILD_ID)-$(OS)-$(ARCH) > release/$(BINARY_NAME)-$(BUILD_ID)-$(OS)-$(ARCH).sha256
 endif
+
+dist-aarch64: 
+	CMAKE_ARGS="$(CMAKE_ARGS) -DLLAMA_NATIVE=off" GRPC_BACKENDS="backend-assets/grpc/llama-cpp-fallback backend-assets/grpc/llama-cpp-cuda backend-assets/grpc/huggingface backend-assets/grpc/bert-embeddings backend-assets/grpc/llama-ggml backend-assets/grpc/llama-cpp-grpc backend-assets/util/llama-cpp-rpc-server backend-assets/grpc/gpt4all backend-assets/grpc/rwkv backend-assets/grpc/whisper backend-assets/grpc/local-store" \
 	$(MAKE) build
 	mkdir -p release
 # if BUILD_ID is empty, then we don't append it to the binary name